Traditional breeding doesn't work for eliminating toxins like gossypol because completely eliminating them leaves plants vulnerable. In 1954, researchers found a cotton plant that did not make gossypol, and in the early 1960s they crossbred it with the agricultural version of the plant. The seeds of this plant could be roasted and eaten like a nut or crushed and mixed with flour to make high-protein bread. "Farmers realized the plants were being chewed up by insects and didn't want to grow them," says Rathore.

The advantage of Rathore's RNAi approach is its specificity. The seeds are nearly free of toxin, but the rest of the plant has normal levels of gossypol, so it should be hardy. The same holds true for the next two generations of plants grown from the seeds, Rathore says.

Abhaya Dandekar, professor of pomology at the University of California-Davis, who works on transgenic plants, says the gene-silencing technique could be broadly useful. "The only issue is how stable this will be in production," he says. "Environmental conditions and viruses can interfere with the gene-silencing mechanism in plants."

Scheffler says that the Texas A&M researchers face an additional challenge in getting the edible cotton plants onto farms. The cotton plants used for research like Rathore's are not the same ones in agricultural use. Traditional breeding will be needed to transfer the trait to the agricultural cotton species. Researchers will then need to test the resulting plants for the stability of the RNAi.